Strategies for treatment of tibial plafond fractures have changed since classic open reduction and fixation with large plates and screws was shown to be associated with a high rate of early soft-tissue complications, including skin necrosis, superficial infection, osteomyelitis, and even amputation
1-4 . The use of external fixation with or without delayed plate fixation after soft-tissue recovery is now emphasized, and the prevalence of catastrophic soft-tissue complications has decreased
5-16 . Early clinical outcomes are comparable with those achieved with the previous techniques
16 ; however, published studies have involved only relatively short follow-up periods and have concentrated on details of treatment, healing rates, and complications
5-18 .
Since 1988, we have treated tibial plafond fractures with a uniform technique of external fixation coupled with limited internal fixation of the articular surface with screws. The purpose of this study, performed at a minimum of five years after injury, was to determine the impact of these fractures on ankle pain and function, to assess the development of post-traumatic osteoarthritis, and to identify factors related to clinical and radiographic outcome. In addition, we sought to compare sequential outcome measures over time to gain further insight into the postinjury recovery of these joints.
This study was performed with the approval of our institution's Human Subjects Review Board. All study patients were contacted in accordance with guidelines set up by the Board and all signed a consent form approved by the Board.
Subjects
Between January 1988 and December 1994, fifty-six ankles in fifty-two patients were treated for a tibial plafond fracture (AO/OTA type B3, C1, C2, or C3) at the University of Iowa Hospitals and Clinics or WakeMed Center, Raleigh, North Carolina. All ankles were treated with a uniform technique, previously described by one of us (J.L.M.) and colleagues
6,11 , which consisted of the application of a monolateral hinged transarticular external fixator coupled with screws to fix the articular surface. The fixator spanned two pins distal to the ankle, one in the talus, one in the calcaneus, and two in the medial aspect of the tibia proximal to the fracture. The injury and postoperative radiographs of one patient (Case 4L) are shown in
Figs. 1-A, 1-B ,
1-C, 1-D ,
1-E, 1-F ,
1-G, and 1-H . The fifty-six fractures included all of the fractures treated at the University of Iowa Hospitals and Clinics and all of those treated by one of the investigators (D.R.D.) at WakeMed Center during the study period.
Most of the patients had not been followed for several years. To locate them, a search of the records from the hospitals, the Department of Motor Vehicles, and the state Department of Corrections as well as an Internet search was performed according to the techniques described by Smith and Watts
19 . Patients were contacted by either telephone or letter and were invited to participate in the study.
Three patients had died during the follow-up period, and thirteen either had been lost to follow-up or refused to participate. Early in the study period, three ankles were fused within six months after the fracture because it was thought that the articular injury was so comminuted it could not be reduced and the ankle would not provide useful function. Two additional ankles were fused because of secondary degenerative arthritis and pain at nineteen and twenty-four months after the injury. These five patients were not studied further.
Thirty-one patients (70% of the forty-four available patients and 60% of the fifty-two originally treated patients) with thirty-five involved ankles had radiographic and clinical follow-up at a minimum of five years (average, 79.6 months; range, sixty to 146 months) after the injury. Twenty-nine patients returned to the clinic specifically for the purposes of this study. These patients included fourteen patients who had been part of a previous study and outcome measures had been recorded two to four years after the injury for nine of those patients
11 . Two patients could not be located at the time of the present study, but follow-up radiographs and clinical notes had been made, and the patients had completed some of the questionnaire material, at a minimum of five years after the injury. However, as these two patients had not been recently examined and had not completed all of the study questionnaires, they were not included in the statistical analysis.
Basic demographic data are summarized in the Appendix. There were twenty-three men and eight women. Eighteen fractures involved the right ankle, and seventeen involved the left. The majority of fractures were sustained in motor-vehicle accidents (seventeen automobile and two motorcycle accidents). Eleven injuries resulted from a fall and one occurred from an auger injury. There were twelve open fractures: two were Gustilo type I, one was type II, and nine were type IIIA
20,21 . Eighteen fractures were associated with injuries of the ipsilateral lower extremity. Sixteen patients were manual laborers, twelve were employed in an office, and one patient was retired.
Clinical Evaluation
A physical examination was performed by one of us (D.P.W. or D.R.D.). Gait was assessed and was recorded as either antalgic or nonantalgic. Limb length was assessed with use of the block method. Combined maximum ankle and foot dorsiflexion and plantar flexion were measured with the foot in the weight-bearing position with a goniometer placed flat on the floor and along the tibial shaft. Subtalar motion was assessed with use of the technique described by McMaster
22 . Motion of both the ankle and the subtalar joint was measured bilaterally. Neurologic examination was performed to assess sensory abnormalities to light touch, and if any were present, they were described according to their distribution of cutaneous nerves. Clawing of the toes was noted to be present or absent. The height and weight of the patient were recorded.
Questionnaires
The patients were asked to complete four questionnaires. On a general questionnaire, they were asked questions that were possibly relevant to the outcome of the fracture but were not included in the other questionnaires. They were asked whether they had had any surgical procedures on the ankle since the time of the initial treatment; about the use of pain medication for the ankle, ambulatory aids, and shoe-wear modifications; whether they were able to run; the time to maximal improvement; their satisfaction with the outcome; and their work status.
The patients also completed the Short Form-36 (SF-36), a general health survey, to assess the effect of the fracture on physical and emotional health and on health-related quality of life more than five years after the injury.
The Iowa Ankle Score, as described by Merchant and Dietz
23 , was calculated as well. This score, which is based on a 100-point scale, incorporates the patient's self-assessment of pain and function as well as data from the physical examination with regard to the range of motion, gait, and deformity. It provides a joint-specific measure of outcome. It had been previously utilized for nine of the patients at two to four years after the injury
11 .
Radiographics Analysis
Injury radiographs were reviewed and fractures were classified according to the AO/OTA System
25 . The severity of the injury was assigned a rank of 1 to 10, with 1 representing the most severe and 10 representing the least severe. Reduction, as seen on radiographs, was also assigned a rank of 1 to 10, with 1 being the worst reduction and 10 representing an anatomic reduction. This is a variation of the rank-order methodology that has been previously utilized to assess injury severity and quality of reduction
26 . Reduction was also assessed with use of a modification of the criteria of Burwell and Charnley
27 . With their system, reduction is classified as good, fair, or poor on the basis of the displacement of the fracture fragments (medial, lateral, and posterior malleoli, anterior lip, and central fragment), the talus (tilt or subluxation), and mortise widening; all parameters are measured in millimeters except for tilt, which is measured in degrees
11 .
Alignment was assessed, on standing anteroposterior, lateral, and mortise radiographs of the ankle made at the time of follow-up, by measuring the angle created by the intersection of the subchondral line of the plafond and a line drawn up the center of the tibial shaft. Ninety degrees was considered normal, and deviations of >5° were recorded as either varus, valgus, anterior, or posterior angulation.
Arthrosis was graded as well: Grade 0 indicated no evidence of arthrosis; grade 1, small spurs but no joint-space narrowing; grade 2, osteophytes and some joint-space narrowing; and grade 3, complete loss of the joint space
11 . For nine patients, the arthrosis seen on current radiographs was compared with that seen on previous radiographs made between two and four years after the injury.
Statistical Analysis
The Wilcoxon signed-rank test for nonparametric distribution of dependent ordinal variables was used to compare the earlier Iowa Ankle Scores for the nine patients who had been part of the previous study with their scores in the current study. To examine the relationships between risk-factor variables and outcome variables, the Kendall rank correlation was used to adjust for tied ranks resulting from identical values. The level of significance was set at p < 0.05, meaning that the probability of testing the null hypothesis that the Wilcoxon signed rank or the Kendall rank correlation coefficient equals zero is <0.05. To account for the multiple variables being tested, a multivariate adjustment was performed with use of Cochran-Mantel-Haenszel statistics, with the level of significance again set at p < 0.05. The Spearman correlation coefficient was used to assess the relationship between the injury and reduction ranks and arthrosis. The group t test was used to compare these continuous variables with the level of significance set at p < 0.05.
Clinical Evaluation
The data obtained at the physical examination are presented in detail in the Appendix. The combined ankle-foot range of motion averaged 10° of dorsiflexion (range, -3° to 30°) and 29° of plantar flexion (range, 10° to 55°). On the average, the total arc of ankle motion on the involved side (39°) was 62% of that on the contralateral side (63°). The subtalar range of motion averaged 10.2° of inversion and 9.6° of eversion, with the mean arc (20°) averaging 77% of that on the contralateral side (26°). Five patients had evidence of clawing of the toes. Eleven patients had an antalgic gait, and eleven had shortness of the limb ranging from 0.5 to 1.5 cm (mean, 0.86 cm). Three patients had decreased sensation in the distribution of the deep peroneal nerve, and one patient, who had sustained a lumbar burst fracture, had partial paraparesis and walked with a bilateral ankle-foot orthosis.
Questionnaires
The average age at the time of follow-up was forty-four years. Two patients had had prominent screws removed. Other than the five ankles that had undergone fusion and are not a part of this report, no other ankle had a surgical procedure after treatment of the fracture. Seven patients used orthotic inserts, four wore a leather ankle-lacer, and the patient with partial paraparesis used a bilateral ankle-foot orthosis. At the time of follow-up, twelve patients were laborers, eight were office workers, four were retired, and five considered themselves to be disabled. Of twenty-three patients who had changed occupation since the fracture, fourteen did so because of the ankle injury. The majority of the patients had some limitation with regard to recreational activities, with an inability to run being the most common complaint (twenty-seven of the thirty-one patients). No pain medication was used for fifteen ankles, acetaminophen or nonsteroidal anti-inflammatory drugs were occasionally used for nineteen ankles, and narcotics were occasionally used for one ankle.
There was a wide range of functional recovery from the ankle fracture and a wide range of severity of ankle pain, as reflected by the Iowa Ankle Scores, which ranged from 28 to 96 points (average, 78 points) (see Appendix).
The scores on the SF-36 (
Fig. 2 ) and Ankle Osteoarthritis Scale (
Fig. 3 ) were compared with known normative data for age-matched controls
24,28 . The SF-36 subscale scores for physical function (p < 0.001), physical role (p = 0.002), and bodily pain (p = 0.002) were significantly decreased, but those for general health, vitality, social function, emotional role, and mental health were not significantly different from the control values. As seen in
Figure 3 , the pain and disability scores on the Ankle Osteoarthritis Scale were markedly worse then those for the age-matched controls, a finding that further substantiates the long-term effect of this injury on ankle pain and function. Comparison of
Figures 2 and
3 reveals that the joint-specific measure (Ankle Osteoarthritis Scale) showed a much more dramatic negative effect of this injury than did the general health-status measure (SF-36).
Despite these measured difficulties, the patients were generally satisfied with the results of treatment and perceived that they had improved for a long time after the injury. Fifteen ankles were rated by the patient as excellent; ten, as good; seven, as fair; and one, as poor. The patients reported that the time to maximal improvement was between nine months and five years (average, 2.4 years). The Iowa Ankle Scores of the nine patients for whom they had been recorded in the earlier study
11 confirmed this patient perception of prolonged improvement, as all showed improvement over time. The average score for these nine patients was 86 points at ninety-two months after the injury compared with 67 points between twenty-four and forty-eight months. This difference was significant (p = 0.004, Wilcoxon signed-rank test).
Radiographic Analysis
The radiographic data at the time of injury and at the time of follow-up are presented in detail in the Appendix. There were ten B3, three C1, ten C2, and twelve C3 injuries. The fracture type was not associated with the clinical outcome measures. According to the criteria of Burwell and Charnley
27 , the quality of the reduction was rated as good in fourteen ankles, fair in fifteen, and poor in six. The degree of osteoarthrosis was grade 0 in three ankles, grade 1 in six, grade 2 in twenty, and grade 3 in six. Comparison of the earlier radiographs of the nine patients in the previously reported study
11 with their current radiographs showed that the osteoarthrosis had increased by one grade in two ankles, had increased by two grades in one, and had remained unchanged in six.
Figures 1-A, 1-B ,
1-C, 1-D ,
1-E, 1-F ,
1-G, and 1-H illustrate degenerative changes present at two years that had progressed by six years after the injury.
Statistical Analysis
With the numbers available, the duration of follow-up, gender, height, weight, education level, work status, associated injuries, and presence of an open fracture had no significant associations with any of the outcome measures. The age of the patient was negatively associated with the SF-36 mental health score (correlation coefficient = -0.27, p = 0.033).
There were more associations between radiographic factors and outcome variables. The AO/OTA fracture type was positively associated with the arthrosis grade (correlation coefficient = 0.38, p = 0.010) but was not associated with any clinical outcome measures. The injury rank was associated with the SF-36 physical function score (correlation coefficient = 0.26, p = 0.043) and the arthrosis grade (correlation coefficient = 0.41, p = 0.003). Both measures of reduction were significantly correlated with outcome variables. The reduction rank correlated with the Iowa Ankle Score (correlation coefficient = 0.35, p = 0.007); the SF-36 physical function (correlation coefficient = 0.29, p = 0.030), bodily pain (correlation coefficient = 0.35, p = 0.009), and social function (correlation coefficient = 0.30, p = 0.032) scores; and the arthrosis grade (correlation coefficient = 0.48, p = 0.0006). The Burwell and Charnley class correlated with the Iowa Ankle Score (correlation coefficient = 0.42, p = 0.003), SF-36 physical function score (correlation coefficient = 0.37, p = 0.01), Ankle Osteoarthritis Scale disability score (correlation coefficient = 0.31, p = 0.03), and arthrosis grade (correlation coefficient = 0.54, p = 0.0005). Angular malalignment in varus or valgus correlated with the SF-36 physical role score (correlation coefficient = 0.32, p = 0.027), and anterior or posterior angulation correlated with the SF-36 physical role (correlation coefficient = 0.34, p = 0.027) and social function (correlation coefficient = 0.39, p = 0.011) scores.
Use of the Spearman correlation showed a strong correlation between the injury rank and the reduction rank (correlation coefficient = 0.63 [95% confidence interval = 0.33 to 0.93], p = 0.0003). The arthrosis grade had a strong correlation with both the injury severity rank (correlation coefficient = 0.62, p = 0.0005) and the reduction rank (correlation coefficient = 0.65, p = 0.0002). If covariation with the reduction rank, age, duration of follow-up, height, and weight are removed, the partial correlation between injury severity rank and arthrosis has a magnitude of 0.34 (p = 0.079). If, instead, covariation with the injury rank and the same other factors is removed, the partial correlation between reduction rank and arthrosis grade has a magnitude of 0.42 (p = 0.027).
As an adjustment for multiple testing, we performed another analysis of the data with only correlations with an absolute magnitude of 0.46, indicating that the risk factor accounted for 20% of the ranked variation in an outcome measure, accepted as significant. When these stricter criteria for significance were used, the only significant relationships between risk factors and outcome measures were those between the two measures of reduction quality (reduction rank and Burwell and Charnley class) and the arthrosis grade.
For multivariate analysis, we performed a multiple regression controlling for the risk factors of age, duration of follow-up, height, and weight and then tested correlations with the Cochran-Mantel-Haenszel statistics. After this analysis, the associations between anteroposterior angulation and the SF-36 physical function and social function scores and the association between the injury and reduction ranks and the arthrosis grade remained significant. The presence of arthrosis on radiographs had weakly positive correlations with the clinical outcome measures (Iowa Ankle Score, correlation coefficient = -0.23; Ankle Osteoarthritis Scale pain score, correlation coefficient = 0.37; Ankle Osteoarthritis Scale disability score, correlation coefficient = 0.30; SF-36 physical function score, correlation coefficient = -0.32; SF-36 physical role score, correlation coefficient = -0.04; and SF-36 bodily pain score, correlation coefficient = -0.37).
The outcomes more than five years after treatment of tibial plafond fractures have not been widely reported
7,29-32 ; in particular, there have been few reports with detailed functional assessment and use of current outcome measures. To our knowledge, the longest duration of follow-up (average, ten years) in the English-language literature was in the study by Etter and Ganz
29 . In that report, forty-one patients who had a type-I, II, or III fracture according to the criteria described by Ruedi and Allgower
32 were treated with open reduction and internal fixation. Of the twenty type-III fractures (49% of the group), nine (45%) were followed by the development of moderate arthrosis and six (30%) were followed by the development of severe arthrosis, indicating a long-term impact of these injuries on the ankle joint.
With use of validated outcome measures, we also found that tibial plafond fractures have a long-lasting negative effect on ankle function, work, recreation, and health-related quality of life. Patients frequently used pain medication, changed jobs, and were unable to participate in all recreational activities. Furthermore, if the five patients who had an ankle arthrodesis are considered to have had a poor outcome because the fracture led to arthrodesis, the overall results are even less favorable.
Arthrosis was seen in the majority of the ankles, but the effect of this radiographic finding on clinical outcome was not clear. Some joint-space narrowing was evident in 57% (twenty) of the thirty-five ankles, and severe arthrosis was seen in 17% (six) of the thirty-five. However, the presence of arthrosis had only weak correlations with clinical outcome, as measured with the Iowa Ankle Score, the Ankle Osteoarthritis Scale, and the SF-36. Etter and Ganz also noted this lack of correlation between clinical outcome and radiographic signs of arthrosis
29 .
Despite the long-lasting negative impact of these injuries, few ankles required late arthrodesis (two) or any subsequent ankle surgery, most patients were working, and twenty-five of the thirty-five ankles were rated as good or excellent by the patients. Although our assessment techniques are imperfect and our data are incomplete, the results suggest that the majority of patients have prolonged recovery. The patients reported improvement for an average of 2.4 years, and the nine patients with sequential ankle scores all had improvement at the time of the longer follow-up. In 1973, Ruedi
31 presented sequential follow-up results in a group of patients with a tibial plafond fracture, and he also noted that most patients had an improved or stable outcome after the longer follow-up interval and that only a few had deterioration. On the basis of this optimistic prognosis for prolonged improvement, we urge caution in recommending early reconstructive procedures.
With multivariate analysis, the severity of the injury and the quality of the reduction did not correlate with the clinical outcomes measured with the Iowa Ankle Score, Ankle Osteoarthritis Scale, or SF-36 and did not account for more than a small percentage of the variance of any of these measures. Not surprisingly, the quality of the reduction achieved was tightly linked with the severity of the original injury (correlation coefficient = 0.63, p = 0.0003). We interpreted these data to mean that the reasons for the variability in patient outcomes at more than five years after tibial plafond fractures are multifactorial and there may be factors that we were unable to assess or measure. Our small sample size may have prevented us from identifying potentially meaningful associations. In addition, the five patients who underwent an arthrodesis were not included in the analysis. Considering that they had severe injuries, poor reductions, and poor outcomes, inclusion of these patients may have resulted in more significant associations. Finally, with longer follow-up of these patients, a stronger association between arthrosis and clinical outcome may become apparent.
The severity of the injury has been shown to be predictive of short-term outcome
5,15,16,33 . Kellam and Waddell described better clinical outcomes after lower-energy rotational fractures than after the more severe axial loading fractures
33 . Barbieri et al.
5 and Wyrsch et al.
16 noted significantly worse outcomes in patients with AO/OTA type-C2 and C3 fracture patterns. However, this association was not seen in the current study, which included mostly higher-energy fractures with longer follow-up. Distinguishing between lower-energy, mostly rotational fractures and higher-energy axial compression fractures is a clear predictor of patient outcome. Our data suggest that, in the group of fractures caused by axial compression, it may not be possible to further stratify injury severity (fracture classification) to predict outcome after longer follow-up, given the problems with our current assessment techniques
34-37 .
Most surgeons believe that articular reduction is critically important, and all of the fractures in the present series were reduced to some extent. However, our measurements of the degree to which we actually reduced the articular surface had only weak associations with clinical outcome. It is possible that our techniques of measuring reduction quality were not accurate for assessing subsequent variations in joint-contact loading. Alternatively, reduction may have less of an effect than was previously thought, or it may be difficult to detect its long-term effect in a relatively small number of patients because that effect is so overshadowed by other factors. This lack of correlation has also been seen in other studies in which modern clinical outcome measures have been used
16,26 .
Although it is difficult to use the data in our study to compare our technique of treatment with other techniques, the need for arthrodesis appears to be similar to that in other studies with long-term follow-up. In our study, an arthrodesis was performed in 13% (five) of the forty ankles for which the outcome was known at a minimum of five years after the injury. When only ankles that required late arthrodesis—that is, after the initial treatment period (more than six months after the fracture)—are considered, the rate drops to 5% (two of thirty-seven ankles). Ruedi and Allgower
32 reported a rate of arthrodesis of 5% (four of seventy-five) at an average of nine years after the injury. Bourne et al.
7 reported a 17% rate of arthrodesis (seven of forty-two) at an average of fifty-three months after plate fixation, with the majority of the fusions performed in patients with a severely comminuted fracture. Ovadia and Beals
30 reported a 12% rate of arthrodesis or replacement at an average of fifty-seven months following a variety of treatments for fractures of the tibial plafond.
One weakness of the current study is the number of patients who were lost to follow-up and the potential impact of that loss on our results. Although every effort was made to contact and recruit all patients, we were able to obtain follow-up data on only 70% (thirty-one) of the forty-four potentially available patients.
In summary, after five to twelve years of follow-up, tibial plafond fractures were found to have a significant effect on general health and a greater effect on the involved ankle as measured by the development of arthrosis and an ankle-specific outcome instrument. Despite these difficulties, the rate of late arthrodesis is low and many patients continue to improve for many years after the injury. On multivariate analysis, injury severity and reduction quality were found to correlate with the development of arthrosis, but radiographic signs of arthrosis had only a weak correlation with clinical outcome.
Tables showing demographic data and detailed clinical and radiographic results are available with the electronic versions of this article, on our web site at www.jbjs.org (go to the article citation and click on "Supplementary Material"), and on our quarterly CD-ROM (call our subscription department, at 781-449-9780, to order the CD-ROM).